Ready-Made AI Agents: Catalog & Comparison

Not every agent needs to be built from scratch. Dozens of ready-made tools already edit code, execute tasks autonomously, and integrate with IDEs.
This section is a guide to existing AI agents: how they differ, which to choose, and how to connect them to local models.

← AI agents · Skills for agents → · 🇷🇺 Русский

Contents

1. How to Read This Table
2. Comparison Table
3. CLI Agents
4. IDE Plugins
5. AI-native IDEs
6. Autonomous Agents
7. Orchestrator Agents
8. Which Agent to Choose
9. What’s Next

1. How to Read This Table

All agents are divided into 5 categories by usage style:

Local Model Integration Levels:

• Full — agent works with any model via Ollama / OpenAI-compatible API
• Limited — local models connect, but not all features available
• Cloud Only — agent requires a proprietary model

2. Comparison Table

Agent	Type	Local	Install	Stars
Aider	CLI	Full	pip install aider-install	25K+
Codex CLI	CLI	Cloud	npm install -g @openai/codex	15K+
Claude Code	CLI	Cloud	npm install -g @anthropic-ai/claude-code	10K+
Codebuff	CLI	Full	brew install codebuff	1K+
Continue	IDE	Full	Install from VS Code Marketplace	25K+
Cline	IDE	Full	Install from VS Code Marketplace	15K+
Cody	IDE	Limited	Install from VS Code Marketplace	5K+
Cursor	IDE	Limited	Download from website	—
Windsurf	IDE	Limited	Download from website	—
OpenHands	Autonomous	Full	docker pull	40K+
AutoCodeRover	Autonomous	Full	pip install auto-code-rover	3K+
Sisyphus / OhMyOpenAgent	Orchestrator	Full	bunx oh-my-openagent install	66K+
Copilot	IDE	Cloud	Install from Marketplace	—

3. CLI Agents

Work in the terminal, edit files, work with git. Perfect for those who don’t want to leave the console.

Aider

GitHub → · 25K+ · Full local model support

What it does: Edits code in a git repository via text commands. Makes edits itself, runs tests, creates commits.

Key feature: Works with a pair of models — one architect (plans), one editor (writes code). Supports “lazy” mode for cheap models.

pip install aider-install
aider-install

# Run with local model
aider --model ollama/qwen2.5-coder:7b

# Architect mode (2 models)
aider --model ollama/qwen2.5-coder:7b --editor-model ollama/qwen3:4b

When to choose: Daily terminal development. Refactoring. Multi-file changes.

Codex CLI

GitHub → · 15K+ · Cloud only (OpenAI)

What it does: Official CLI agent from OpenAI. Runs code in sandbox environment, edits files.

Key feature: Native OpenAI API integration, sandbox isolation, extension support.

npm install -g @openai/codex
codex "add sorting to task list"

Claude Code

GitHub → · 10K+ · Cloud only (Anthropic)

What it does: CLI agent from Anthropic. Works with Claude.

Key feature: Strong safety, CLAUDE.md instruction support, large context handling.

Codebuff

GitHub → · 1K+ · Full local model support

What it does: Lightweight CLI agent for file operations.

Key feature: Simplicity. One command — one change. No sandboxes.

4. IDE Plugins

Extensions for VS Code, JetBrains, and other editors. Add AI features directly to the interface.

Continue

GitHub → · 25K+ · Full local model support

What it does: Free Copilot with local model. Tab autocomplete, chat, inline editing, @codebase.

Key feature: Only IDE plugin with full Ollama + MCP server + custom rules support. Configured via JSON.

{
  "models": [{
    "title": "Local Coder",
    "provider": "ollama",
    "model": "qwen2.5-coder:7b",
    "apiBase": "http://localhost:11434"
  }],
  "tabAutocompleteModel": {
    "provider": "ollama",
    "model": "qwen2.5-coder:1.5b"
  }
}

When to choose: If you’re in VS Code and want a free Copilot alternative with local models.

Cline

GitHub → · 15K+ · Full local model support

What it does: Autonomous agent inside VS Code. Reads files, writes code, runs commands itself.

Key feature: Can create and execute plans autonomously. Supports MCP servers.

When to choose: If you need an autonomous agent inside the editor, not just autocomplete.

Cody

GitHub → · 5K+ · Limited local model support

What it does: AI assistant from Sourcegraph. Chat, autocomplete, commands.

Key feature: Sourcegraph integration for code search. Strong contextual awareness.

5. AI-native IDEs

Full-featured IDEs with AI built into every corner of the interface.

Cursor

Website → · Limited local model support

What it does: VS Code fork with built-in AI at every level. Tab, chat, agent, terminal.

Key feature: Most polished AI experience. Agent mode plans and executes multi-step tasks autonomously.

With local models: Cursor can connect custom models via OpenAI-compatible API. Not all features work — some are tied to Cursor’s proprietary models.

Windsurf

Website → · Limited local model support

What it does: IDE from Codeium. Cascade — mode where AI sees the entire project.

Key feature: “Cascade” — agent that understands the whole codebase and can execute tasks autonomously.

6. Autonomous Agents

Run in isolated environments and execute tasks from start to finish without human involvement.

OpenHands

GitHub → · 40K+ · Full local model support

What it does: Autonomous software engineer in a Docker container. Creates plan → writes code → runs → fixes bugs → makes PR.

Key feature: Full isolation (Docker), web browser support, local model support.

docker run -it --rm \
  -e LLM_MODEL=qwen2.5-coder:7b \
  -e LLM_API_KEY=ollama \
  -e LLM_BASE_URL=http://host.docker.internal:11434 \
  -v $(pwd)/workspace:/workspace \
  ghcr.io/all-hands-ai/openhands:latest

When to choose: If you need “set it and forget it”. Automatic task execution, bug fixing, PR creation.

AutoCodeRover

GitHub → · 3K+ · Full local model support

What it does: Autonomous agent for bug fixing and feature implementation.

Key feature: Focused on code repair. Can explore code, localize bugs, propose fixes.

7. Orchestrator Agents

Manage teams of specialized subagents — don’t write code themselves, but coordinate those who do.

Sisyphus (OhMyOpenAgent)

GitHub → · 66K+ · Full local model support

What it does: Plugin for OpenCode / Codex CLI adding multi-agent architecture. 11 agents, 54+ lifecycle hooks, 5 built-in MCPs, Team Mode. Exists in two editions:

Key feature: Only system where each agent uses its own model, optimal for its task, while the orchestrator (Atlas) manages their work like a conductor.

Agent Team

How it works (real architecture from repo):

User
    │
    ▼
┌─────────────────────────────────────────────────────┐
│  Planning Layer (Prometheus + Metis + Momus + Oracle)│
│  ┌──────────┐   ┌──────────┐   ┌──────┐  ┌──────┐  │
│  │Prometheus├──►│  Metis   │   │Momus │  │Oracle│  │
│  │(plans)   │   │(analyzes)│   │(review)│  │(review)│  │
│  └────┬─────┘   └──────────┘   └──────┘  └──────┘  │
│       │          ▲               ▲         ▲        │
│       └──────────┴───────┬───────┘         │        │
│                          │                 │        │
│                     ┌────┴─────┐            │        │
│                     │ .omo/plans/*.md       │        │
│                     └──────────┘            │        │
└─────────────────────────────────────────────┼────────┘
                                              │
    /start-work                               │
    ▼                                         │
┌─────────────────────────────────────────────┼────────┐
│  Execution Layer (Atlas)                    │        │
│  ┌────────────────────────────────────────┐ │        │
│  │  Atlas (conductor) reads plan          │ │        │
│  │  and delegates tasks:                  │ │        │
│  │                                        │ │        │
│  │  task(category="deep") ───► Junior     │ │        │
│  │  task(subagent_type="oracle") ──► Oracle │        │
│  │  task(category="visual-engineering")   │ │        │
│  │      ───► Frontend                     │ │        │
│  │  explore ───► Explore agent            │ │        │
│  └────────────────────────────────────────┘ │        │
└─────────────────────────────────────────────┼────────┘
                                              │
    Result ◄──────────────────────────────────┘

When to choose: If you want a ready-made multi-agent team in your editor — 11 agents with different roles and models, no manual setup needed.

This handbook was written by this same team.

How to Install

# Ultimate Edition (OpenCode)
bunx oh-my-openagent install

# Light Edition (Codex CLI)
npx lazycodex-ai install

# Both editions at once
bunx oh-my-openagent install --platform=both

Tip: Open the installation guide via an AI agent:

Install and configure oh-my-openagent by following the instructions here:
https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/refs/heads/dev/docs/guide/installation.md

More details — in the separate orchestrators guide.

8. Which Agent to Choose

9. What’s Next

| If you want to | Go to | |—————|——-| | Understand how orchestrators with subagents work (Sisyphus, Atlas, Prometheus) | Orchestrators with Subagents → | | Understand how to extend agent capabilities (skills, MCP, plugins) | Skills for Agents → | | Build your own agent from scratch | AI Agent Architecture | | Choose a framework for a custom agent | Frameworks | | Set up a local model for an agent | Ollama for Agents | | Return to the agents section | README.md | —

In section: architecture · evaluation · frameworks · memory · multi-agent · ollama-for-agents · orchestrators · patterns · prompting · ready-made · safety · skills
Related sections: Zero Level · Local Models · Use Cases · Resources
Navigation: ← AI Agents · ↑ Back to main · 🇷🇺 Русский